Abstract
For normal models with \(X \sim N_p(\theta, \sigma^{2} I_{p}), \;\; S^{2} \sim \sigma^{2}\chi^{2}_{k}, \;\mbox{independent}\), we consider the problem of estimating θ under scale invariant squared error loss ||d − θ||2/σ 2, when it is known that the signal-to-noise ratio \({\left\|\theta\right\|}/{\sigma}\) is bounded above by m. Risk analysis is achieved by making use of a conditional risk decomposition and we obtain in particular sufficient conditions for an estimator to dominate either the unbiased estimator δ UB (X) = X, or the maximum likelihood estimator \(\delta_{ML}(X,S^2)\), or both of these benchmark procedures. The given developments bring into play the pivotal role of the boundary Bayes estimator δ BU,0 associated with a prior on (θ,σ 2) such that θ|σ 2 is uniformly distributed on the (boundary) sphere of radius m σ and a non-informative 1/σ 2 prior measure is placed marginally on σ 2. With a series of technical results related to δ BU,0; which relate to particular ratios of confluent hypergeometric functions; we show that, whenever \(m \leq \sqrt{p}\) and p ≥ 2, δ BU,0 dominates both δ UB and δ ML . The finding can be viewed as both a multivariate extension of p = 1 result due to Kubokawa (Sankhyā 67:499–525, 2005) and an unknown variance extension of a similar dominance finding due to Marchand and Perron (Ann Stat 29:1078–1093, 2001). Various other dominance results are obtained, illustrations are provided and commented upon. In particular, for \(m \leq \sqrt{p/2}\), a wide class of Bayes estimators, which include priors where θ|σ 2 is uniformly distributed on the ball of radius m σ centered at the origin, are shown to dominate δ UB .
Similar content being viewed by others
References
Abramowitz, M. and Stegun, I. (1964). Handbook of mathematical functions with formulas, graphs, and mathematical tables. Dover, New York.
Eaton, M.L. (1989). Group invariance applications in statistics. Regional Conference Series in Probability and Statistics, vol. 1. Institute of Mathematical Statistics, Hayward, California.
Efromovich, S.Y. (1999). Nonparametric curve estimation: Methods, theory and applications. Springer, New York.
Fourdrinier, D. and Marchand, É. (2010). On Bayes estimators with uniform priors on spheres and their comparative performance with maximum likelihood estimators for estimating bounded multivariate normal means. J. Multivariate Anal., 101, 1390–1399.
Giri, N.C. (2004). Multivariate Statistical Analysis. Marcel Dekker, 2nd edition, revised and expanded.
Hartigan, J. (2004). Uniform priors on convex sets improve risk. Statist. Probab. Lett., 67, 285–288.
Johnstone, I. (2011). Gaussian estimation: sequence and wavelet models, Unpublished manuscript.
Joshi, C.M. and Bissu, S.K. (1996). Inequalities for some special functions. J. Comput. Appl. Math., 69, 251–259.
Kariya, T., Giri, N., and Perron, F. (1988). Invariant estimation of a mean vector μ of N(μ, Σ) with μ′ Σ − 1 μ = 1, Σ − 1/2 μ = C or Σ = δ 2 μ′ μI. J. Multivariate Anal., 27, 270–283.
Kokologiannaki, C.G. (2012). Bounds for functions involving ratios of Bessel functions. J. Math. Anal. Appl., 385, 737–742.
Kortbi, O. and Marchand, É. (2012). Truncated linear estimation of a bounded multivariate normal mean, J. Statist. Plann. Inference, 142, 2607–2618.
Kubokawa, T. (1994). A unified approach to improving on equivariant estimators. Ann. Statist., 22, 290–299.
Kubokawa, T. (2005). Estimation of a mean of a normal distribution with a bounded coefficient of variation. Sankhyā: The Indian Journal of Statistics, 67, 499–525.
Lehmann, E.L. and Casella, G. (1998). Theory of Point Estimation. Springer-Verlag, 2nd edition.
Mandelkern, M. (2002). Setting confidence intervals for bounded parameters with discussion. Statist. Sci., 17, 149–172.
Marchand, É., Jafari Jozani, M. and Tripathi, Y.M. (2012). On the inadmissibility of various estimators of normal quantiles and on applications to two-sample problems with additional information. Contemporary Developments in Bayesian analysis and Statistical Decision Theory: A Festschrift for William E. Strawderman, Institute of Mathematical Statistics Volume Series, 8, 104–116.
Marchand, É. and Perron, F. (2001). Improving on the mle of a bounded normal mean. Ann. Statist., 29, 1078–1093.
Marchand, É. and Strawderman, W.E. (2004). Estimation in restricted parameter spaces: A review. Festschrift for Herman Rubin, IMS Lecture Notes-Monograph Series, 45, pp. 21–44.
Marchand, É., and Strawderman, W.E. (2005). On improving on the minimum risk equivariant estimator of a location parameter which is constrained to an interval or a half-interval. Ann. Inst. Statist. Math., 57, 129–143.
Marchand, É. and Strawderman, W.E. (2006). On the behaviour of Bayesian credible intervals for some restricted parameter space problems. Recent Developments in Nonparametric Inference and Probability: A Festschrift for Micheal Woodroofe, IMS Lecture Notes-Monograph Series, 50, pp. 112–126.
Marchand, É., and Strawderman, W.E. (2012a). On Bayesian credible sets in restricted parameter space problems and lower bounds for frequentist coverage. (arXiv: 1208.002802)
Marchand, É., and Strawderman, W.E. (2012b). A unified minimax result for restricted parameter spaces. Bernoulli, 18, 635–643.
Marchand, É., Strawderman, W.E., Bosa, K., and Lmoudden, A. (2008). On the frequentist coverage of Bayesian credible intervals for lower bounded means. Electron. J. Stat., 2, 1028–1042.
Moors, J.J.A. (1985). Estimation in Truncated Parameter Spaces. Ph.D. thesis, Tilburg University.
Perron, F. and Giri, N.C. (1990). On the best equivariant estimator of mean of a multivariate normal population. J. Multivariate Anal., 32, 1–16.
Roe, B. and Woodroofe, M. (2000). Setting confidence belts. Phys. Rev. D, 63, 013009/01–09.
Watson, G.S. (1983). Statistics on spheres. John Wiley, New York.
Zhang, T. and Woodroofe, M. (2003). Credible and confidence sets for restricted parameter spaces. J. Statist. Plann. Inference, 115, 479–490.
Author information
Authors and Affiliations
Corresponding author
Rights and permissions
About this article
Cite this article
Kortbi, O., Marchand, É. Estimating a multivariate normal mean with a bounded signal to noise ratio under scaled squared error loss. Sankhya A 75, 277–299 (2013). https://doi.org/10.1007/s13171-013-0028-x
Received:
Revised:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s13171-013-0028-x
Keywords and phrases
- Bayes estimators
- coefficient of variation
- confluent hypergeometric functions
- dominance
- estimation
- maximum likelihood
- multivariate normal
- restricted parameter
- signal-to-noise ratio
- squared error loss